Development of actuator fault diagnosis algorithms for hexacopters

Abstract

Actuator faults are one of the common and devasting causes of unmanned aerial vehicles (UAV) accidents and failures, which hinders UAVs’ wider applications due to the direct threat it poses to the public. Hence, the motivation for this project is to address the increasing usage of UAVs for various applications and consequently the need to enhance their safety and reliability to achieve greater acceptance in their usage. In this project, the flight data is generated by HITL simulations on the hexacopter platform, which is chosen due to the additional hardware redundancy that is available as compared to quadcopters. These flight data are then cleaned, preprocessed, and resampled before being used to train a suitable machine learning model for the purpose of fault diagnosis. The selection of the machine learning model is based on an extensive review of past and current methodologies in terms of their advantages and limitations of each technique as well as the expected operating conditions under which a hexacopter will fly. This project aims to develop an algorithm to improve the fault diagnosis of actuators on a hexacopter which it has successfully done so by implementing the Online-Sequential Fuzzy Extreme Learning Model (OS-Fuzzy ELM) that achieved a classification accuracy of 84.8% and F1 score of 0.892.